Access to computing devices, such as mobile devices and digital cameras, has led to a proliferation of digital images, including panoramic images. Panoramic images may be connected (e.g., linked) to each other based on their location and orientation data. The locations at which the panoramic images were captured are typically derived from the global position systems (“GPS”) on the computing devices which captured the images. However, these systems are often inaccurate, and can provide location data which is off by a significant distance, e.g., a few meters or more. Panoramic images may thus be improperly connected together based on the inaccurate data.
When transitioning between connected neighboring panoramic images, the orientations of the two images may be different; this may result in the user being presented with a jarring and confusing experience as they transition between panoramic images. While automatic computer vision based techniques can be used to orient panoramic images relative to each other based on visually similar image data, this can consume significant amounts of processing resources and the results are very often of low quality.